OpenAI Widens Decisions API to All Users, Chasing TypeSafe’s $7.6 Billion Jev Market
Updated
Updated · TechCrunch · Oct 22
OpenAI Widens Decisions API to All Users, Chasing TypeSafe’s $7.6 Billion Jev Market
3 articles · Updated · TechCrunch · Oct 22
Summary
Thursday’s rollout made OpenAI’s Decisions API broadly available after its September 29 debut, pushing the company into the fast, low-cost “decision model” market pioneered by TypeSafe’s Jev.
100 to 180 milliseconds is OpenAI’s target latency, down from one to three seconds, after the team reconfigured its Luna model for faster, cheaper inference rather than training a separate classifier from scratch.
Customer trials showed immediate throughput gains: Speak cut action latency from 1.4 seconds to 0.13 seconds, while Box, Datadog and Ramp used the API to replace slower LLM-based workflows.
OpenAI also added voice and image support and shifted inference resources geographically to reduce delays, betting multimodal flexibility can offset weaker public benchmark standings.
Benchmark pressure remains intense: Liquid’s D1 Ultra and Base Lab’s Elephant-1 lead key indexes, while TypeSafe’s latest Jev ranks seventh and third there and remains the most queried decision model on OpenRouter.